Apple's GPT rival, AI slop & more (September 29, 2025)

Good morning,

From Apple’s internal chatbot tests to global governance clashes, today’s stories highlight how far AI is pushing boundaries, for better and for worse. We trace shifts in oversight, patent innovation, corporate risk, and bold new model launches.

Let’s dig in👇

💰 Money Moves

📰 Apple’s internal Chat GPT

Apple reportedly began internally testing “Veritas” as a ChatGPT-style assistant to bolster Siri’s capabilities. The move suggests Apple is closing the AI gap with rivals who already embed chat models deeply into UX. If it’s rolled into iOS or macOS, this could reframe expectations for on-device intelligence.

💰 AI ‘workslop’ costing companies millions

An HBR analysis warned that AI-generated “workslop” is destroying productivity by producing plausible but shallow drafts that require manual correction. Firms lose time and money undoing AI’s half-finished output. Stronger governance, prompt design, and review regimes may be the antidote.

🌐 U.S. rejects international AI oversight at U.N. General Assembly

The U.S. rejected proposals for binding international AI oversight during the U.N. General Assembly. Doing so emphasizes its preference for national sovereignty in AI regulation over supranational control. The decision raises questions about fragmented global standards and the risks of avoiding coordinated guardrails.

🚀 New Launches

📢 NIT Rourkela Patents AI Model to Boost Road Safety Through Vehicle

Researchers at NIT Rourkela patented an AI model to boost vehicle-to-vehicle communication and improve traffic safety. If deployed, this could reduce accident rates via more anticipatory alerts. It also adds momentum for smart infrastructure and autonomous driving synergies.

💡 Intel AI Development

NVIDIA and Intel announced a collaboration to build AI infrastructure and PC products. The alliance targets next-gen AI hardware across data centers and endpoints. With combined strength, they may shift power in the AI stack.

🚀 Deep Seek has released an experimental version of the AI model

DeepSeek launched an experimental “V3.2-Exp” model to explore enhancements in context length and training efficiency. Early adopters are invited to test and influence its evolution. If successful, this experiment could accelerate alternatives to closed models.

🧰 Tools of the Day

→ Integrity - Unified project brain, docs, canvases, and AI chats together
→ Nexa SDK - Run, build, and ship local AI in minutes
→ opencode zen - AI models curated for developers by the opencode team

✍️ Prompt of the Day

"You are my AI adoption lead. Audit our team’s top five workflows, identify where AI adds measurable ROI this quarter, and propose one pilot per workflow. Specify tools, success metrics, owners, data needs, and risks. Deliver a 1-page brief plus a 30-day rollout checklist."

⚡ Quick Hits

→ Walmart CEO warns AI could cost jobs
→ Anthropic's $1.5B settlement may redefine AI licensing and creator obligations.
→ AI actress Tilly Norwood protests deepfake use
→ Revvity launches AI-powered multimodal imaging software for research
→ Sprinklr unveils new AI features for customer experience

🧾 TLDR

AI is evolving fast: Apple is prototyping an internal chatbot while governance and rights issues dominate the headlines. The U.S. rejected binding global oversight even as NIT Rourkela patents safety tech and Anthropic settles with authors. New launches from DeepSeek and an Intel-NVIDIA alliance signal continued technology arms race in capabilities and infrastructure.

Cheers,
David